{ "epoch": 30.0, "eval_loss": 0.03293671831488609, "eval_mem_cpu_alloc_delta": 85377, "eval_mem_cpu_peaked_delta": 33067, "eval_mem_gpu_alloc_delta": 0, "eval_mem_gpu_peaked_delta": 569275904, "eval_runtime": 4.0268, "eval_samples": 23, "eval_samples_per_second": 5.712, "init_mem_cpu_alloc_delta": 1868923, "init_mem_cpu_peaked_delta": 18306, "init_mem_gpu_alloc_delta": 1370738688, "init_mem_gpu_peaked_delta": 0, "perplexity": 1.0334851364909519, "train_mem_cpu_alloc_delta": 480835, "train_mem_cpu_peaked_delta": 329934, "train_mem_gpu_alloc_delta": 4036706816, "train_mem_gpu_peaked_delta": 8663361536, "train_runtime": 399.4076, "train_samples": 23, "train_samples_per_second": 1.728 }